The output waveform is as follows. It looks quite good. But the measured THD is quite bad. I added the component 'fourier' in analogLib to measure THD. The result is as follows:

Harm:  Absolute        Absolute        Relative        Relative
         Magnitude       Phase           Magnitude       Phase
  1:  276.589 mV      50.4506 Deg     0.000 dB        0 Deg <- normalizer
  2:  43.0674 uV      16.6644 Deg    -76.154 dB      -33.7862 Deg
  3:  161.77 uV       125.855 Deg    -64.659 dB       75.4046 Deg
  4:  5.56332 mV      92.6739 Deg    -33.930 dB       42.2232 Deg
  5:  82.87 uV        160.102 Deg    -70.469 dB       109.652 Deg
  6:  4.03262 mV     -10.6745 Deg    -36.725 dB      -61.1252 Deg
  7:  167.665 uV     -40.5823 Deg    -64.348 dB      -91.033 Deg
  8:  37.2955 uV     -177.267 Deg    -77.404 dB      -227.718 Deg
  9:  27.7985 mV      21.0579 Deg    -19.956 dB      -29.3927 Deg
 10:  333.301 uV      87.8913 Deg    -58.380 dB       37.4407 Deg
Total harmonic distortion = 10.354 % (-19.6978 dB).

Is there something wrong with the measurement, it should not be so high. somebody told me that I do not need to care about the even harmonics because " If you do fourier transform of a signal ranging from say, -V to +V ( i.e., differential), you will find that the  output spectrum will be devoid of any even harmonics." . Is that true?
thanks for any advice.

The Fourier analysis is reporting the distortion in the entire waveform, but presumably the circuit that follows the sample & hold is only sensitive to the settled values of the signal. To model this, you should follow your circuit with an ideal sample & hold and apply the Fourier analysis to the output of the ideal sample & hold. You can find ideal sample & holds by going to the Verilog-AMS section of this website.

-Ken

Berti wrote on Jan 20^th, 2009, 3:29am:



Hi, Berti
I mean the simulated result using Cadence Spectre.

Ken Kundert wrote on Jan 20^th, 2009, 8:44am:



Btw, How to use the Verilog-AMS in spectre?

Hi,

I'd like to calculate the FFT of the SHA output, one point for each cycle after the output settling down. The DFT function in Calculator is easy to use.

Yawei